City-level IP geolocation based on network topology community detection

Abstract

Simple voting rules used in existing IP geolocation method based on network topology heuristic clustering(HC-based) may bring large errors to the geolocation results. In this paper, a city-level IP geolocation method based on network topology community detection is proposed. The method uses the principle that nodes in one network community are usually located in a same metropolitan area network (MAN). First, the modularity is used to measure the community quality and the network topology communities with the highest modularity are obtained through modularity optimization. Second, the communities' locations are gotten by querying geoIP databases. Last, community location where the target IP lies is regarded as the estimated location of target. Experimental results in 3 provinces of China with 9000 nodes show that the proposed method outperforms the HC-based method, which is IP city-level typical geolocation method based on network cluster voting, both in IP geolocation accuracy and recall ratio.